Deciding between man and zone coverage is one of the most critical strategic choices a defensive coordinator must make before each offensive play in American football. While experienced offensive coordinators and quarterbacks often rely on visual cues to identify these defensive schemes, the increasing availability of player tracking data offers a new avenue to uncover and analyze these tactics. A notable example is Amazon’s NFL Next Gen Stats model, which delivers coverage predictions during live broadcasts (see a snapshot of the 2024 Week 12 matchup between the Pittsburgh Steelers and Cleveland Browns). However, these models seem to be trained on plays without pre-snap motion, or at least to the situations before motions (see Amazon), while motion is a crucial element of modern offensive strategies.
Our project takes this model a step further. While we similarly predict man- or zone coverage when the teams are set before snaps, we further leverage the additional information of pre-snap player movements. Using a hidden Markov model (HMM), we model defenders’ trajectories based on hidden states, which represent the offensive players they may be guarding. Incorporating summary statistics of the probabilistic HMM results as features into the existing pre-motion model significantly improves both the AUC and detection accuracy and further allows for evaluating the effectiveness of pre-snap motion in uncovering defensive strategies, providing real-time tactical insights for coaches.
We analyze tracking data from nine weeks of the NFL 2022 season, provided by the NFL Big Data Bowl 2024. Beside the tracking data, we also use information on plays and players. We further considered the corresponding data from PFF that assigned the categories , and representing the different schemes to each play. As it is not properly described what means, we omit every play that is associated with this value. Moreover, we omit plays with more than five offensive linemen and with two quarterbacks. Since we are specifically interested in analyzing pre-snap player movements, we omit plays that did not contain any pre-snap motion. Then, we end up with \(3985\) offensive plays in total, from which the defense played \(2973\) in zone and \(1012\) in man coverage.
To accurately forecast the defensive scheme (man- or zone defense) for every play, we need to create various features derived from the tracking data. In particular, we conducted the following feature engineering steps: We first consider all 11 players on each side of the field and compute features related to the convex hull of the positions of the players. In particular, for defense and offense, we compute the area spanned by the convex hull of all player such as well as the largest \(y\) distance (i.e. the width of the hull) and the largest \(x\) distance (i.e. the length of the hull). In addition, we select the five most relevant players on each side of the field. For offense, we omit the offensive line and the QB, while, for defense, we omit nose tackles, defensive tackles and defensive ends, and select the five defenders that were the closest to the five attackers corresponding to a weighted euclidean distance, putting much more emphasis on the y-axis. Finally, we use their standardized \(x\) and \(y\) coordinates as features and order defensive and offensive players according to their \(y\) coordinates. Additionally, for each of the relevant defenders, we compute distances to the football and their orientation with respect to the quarterback (values taken at event lineset. Rouven: Die Values haben wir doch rausgenommen aus dem finalen Modell oder? Robert: ich denke das sollten wir noch besprechen. Ich habe ohnehin mehrere Modelle gefitted…). Finally, we extract relevant information from play-by-play data, such as quarter, down, yards to go, home and away score and the game clock.
Detailed information can be found in the Appendix.
Our analysis comprises different steps:
We train different models to predict whether the defense plays a man- or zone coverage scheme. In particular, we start with fitting a model only using features available pre-motion. Since the aim of the project is to show the effectiveness of pre-snap motion, this allows us to establish a baseline model for detecting zone vs man coverage without motion.
Since we have a limited dataset available (only 3985 plays), we want to balance the complexity of the model, i.e. we control the number of features available. Using the previously described features, we obtain 67 variables. 6 convex hull related features (3 for offense, 3 for defense), 20 player positions features (10 standardized \(x\) and \(y\) coordinates, 5 for offense, 5 for defense), 30 distance features (10 Euclidian distances, 10 \(x\) distances, and 10 \(y\) distances to the football), 10 orientation features (5 offense, 5 defense) and 6 play by play features (probably better suited for appendix…). Given the small data set this is rather high number of features and therefore we consider two model classes. On the one hand, we fit an elastic net model, which performs implicit feature selection and is able to handle multicollinearity. On the other hand we use an xgboost model, which is able to capture non linear effects (and interactions) and in general also handles collinearity, but needs careful hyperparameter tuning and generally performs better on bigger data sets.